Avoca's Responder Now Books More Than the Typical CSR
Avoca's AI CSR went from about 4 points behind human CSRs in March to level or ahead since July. Shops on the new v3 voice lead the most.
Since AI started answering phones in home services, one question has been central: how do AI CSRs compare with human CSRs?
Avoca has been tracking this closely across about 1 million inbound AI calls a month.
We measure the gap: Avoca's booking rate vs the typical human CSR at the same shop.
AI Catching Up to Humans
When AI voices first came on the scene, they answered mostly overflow, after-hours, and weekend calls, so every call answered and booked was incremental to the business.
As more and more shops trust AI, they are making it their first line of defense, taking every call at all hours.
To that end, AI voice needs to be on par with, and better than, the humans it works alongside.
In March, Avoca's AI CSR trailed human CSRs by about 4 points. The gap narrowed through June.
Avoca drew level in July and has been level or ahead since, peaking at nearly 2 points ahead in August.
New Voice Models and Best Practices Drive the Gain
Two things appear to drive this gap.
First, the voice. This summer Avoca introduced v3, a new generation of our AI voice. Shops began switching to v3 in the middle of July 2026, and by August most of the shops in this study had switched.
V3 voices are more humanlike, natural, and capable. Calls now flow more like talking to a person.
Before July the shops that moved to v3 trailed human CSRs by about 1 to 5 points.
Shops on v3 voices have accelerated the biggest gap: about +4 in October.
Second, best practices. With about a million AI calls a month, we write what works back into every call:
- How to book a job in the fewest number of turns
- When to offer a time
- How to handle a price question
Each call starts from what we learned on the last ones.
Why Avoca Now
We are working to keep growing the gap. Each new voice and each lesson builds on the last. Starting now helps:
- Gains stack. Better voices, playbooks, and data make each next gain bigger.
- Early shops learn what works. Find which calls to send to AI and how to tune it.
- Experience compounds. Shops tuning for months start every improvement ahead.
Methodology
Who is included. ServiceTitan shops only. ~100 shops that had at least 100 Avoca calls and at least 4 human CSRs every month, March to October 2026.
What we measured. Jobs booked per 100 inbound calls handled by an agent. A call counts as booked when a job is attached to it in ServiceTitan.
What Avoca is compared with. The typical (median) human CSR at each shop. Half book more, half book less. Only CSRs with 30 or more calls that month count.
The gap. For each shop and month, Avoca's rate minus the typical human CSR's. We then take the median across shops.
